Table 5.

Accuracy (%) of super-learner models using unsegmented tasks and a combination of segmented tasks and calculated kinesiological features to distinguish controls from PD participants. The numbers in parentheses represent the difference in accuracy between each model and the corresponding standard model, which used data from segmented tasks without incorporating kinesiological features (same row).

	With Segmented Tasks and No Kinesiological Features	With Unsegmented Tasks and No Kinesiological Features	With Segmented Tasks and Kinesiological Features
All PD vs. controls	92.6	89.4 (−3.2)	87.8 (−4.8)
Mild PD (H&Y ≤ 2) vs. controls	89.4	84.3 (−5.1)	88.9 (−0.50)